基于异域数据联邦学习的金属板表面划痕检测

由于真实工业场景下的部分工业缺陷太少,导致深度学习应用在表面缺陷检测困难。针对金属板材表面缺陷检测的真实工业环境中所存在的小样本问题,本文提出了异域数据联邦检测模型,采用轻量级网络MobileNet-YOLOv3,有效利用不同领域的划痕数据,分场景训练检测模型。实验结果表明,通过异域数据结合训练的检测模型拥有一个很好的划痕检测效果,金属板训练的本地模型的mAP为94.95%,比传统数据增广方法训练模型mAP高5%,比YOLOv3的推理速度减少44.8%。

Analysis of Abnormal Characteristics of Regional Crustal Deformation before the Menyuan MS6.4 Earthquake by GPS Continuous Data

In order to study the characteristics of crustal deformation around the epicenter before the 2016 M_S6. 4 Menyuan earthquake,the GPS continuous stations of the period from 2010 to 2016 were selected according to the observation data of the tectonic environment monitoring network in Chinese Mainland. The deformation characteristics of the crust before the earthquake were discussed through inter-station baseline time series analysis and the strain time series analysis in the epicentral region. The results show that a trend turn of the baseline movement state around the epicenter region occurred after 2014,and the movement after 2014 reflects an obvious decreasing trend of compressional deformation.During this period,the stress field energy was in a certain accumulation state. Since the beginning of 2014,the EW-component linear strain and surface strain rate weakened gradually before the earthquake. It shows that there was an obvious deformation deficit at the epicentral area in the past two years,which indicates that the region accumulated a high degree of strain energy before the earthquake. Therefore,there was a significant background change in the area before the earthquake. The results of the study can provide basic research data for understanding the seismogenic process and mechanism of this earthquake.

Regional Variation in Carbon Sequestration Potential of Forest Ecosystems in China

Enhancing forest carbon(C) storage is recognized as one of the most economic and green approaches to offsetting anthropogenic CO_2 emissions. However, experimental evidence for C sequestration potential(C_(sp)) in China's forest ecosystems and its spatial patterns remain unclear, although a deep understanding is essential for policy-makers making decisions on reforestation. Here, we surveyed the literature from 2004 to 2014 to obtain C density data on forest ecosystems in China and used mature forests as a reference to explore C_(sp). The results showed that the C densities of vegetation and soil(0–100 cm) in China's forest ecosystems were about 69.23 Mg C/ha and 116.52 Mg C/ha, respectively. In mature forests, the C_(sp) of vegetation and soil are expected to increase to 129.26 Mg C/ha(87.1%) and 154.39 Mg C/ha(32.4%) in the coming decades, respectively. Moreover, the potential increase of C storage in vegetation(10.81 Pg C) is estimated at approximately twice that of soil(5.01 Pg C). Higher C_(sp) may occur in the subtropical humid regions and policy-makers should pay particular attention to the development of new reforestation strategies for these areas. In addition to soil nutrients and environment, climate was an important factor influencing the spatial patterns of C density in forest ecosystems in China. Interestingly, climate influenced the spatial patterns of vegetation and soil C density via different routes, having a positive effect on vegetation C density and a negative effect on soil C density. This estimation of the potential for increasing forest C storage provided new insights into the vital roles of China's forest ecosystems in future C sequestration. More importantly, our findings emphasize that climate constraints on forest C sequestration should be considered in reforestation strategies in China because the effects of climate were the opposite for spatial patterns of C density in vegetation and soil.Enhancing forest carbon(C) storage is recognized as one of the most economic and green approaches to offsetting anthropogenic CO2 emissions. However, experimental evidence for C sequestration potential(Csp) in China's forest ecosystems and its spatial patterns remain unclear, although a deep understanding is essential for policy-makers making decisions on reforestation. Here, we surveyed the literature from 2004 to 2014 to obtain C density data on forest ecosystems in China and used mature forests as a reference to explore Csp. The results showed that the C densities of vegetation and soil(0–100 cm) in China's forest ecosystems were about 69.23 Mg C/ha and 116.52 Mg C/ha, respectively. In mature forests, the Csp of vegetation and soil are expected to increase to 129.26 Mg C/ha(87.1%) and 154.39 Mg C/ha(32.4%) in the coming decades, respectively. Moreover, the potential increase of C storage in vegetation(10.81 Pg C) is estimated at approximately twice that of soil(5.01 Pg C). Higher Csp may occur in the subtropical humid regions and policy-makers should pay particular attention to the development of new reforestation strategies for these areas. In addition to soil nutrients and environment, climate was an important factor influencing the spatial patterns of C density in forest ecosystems in China. Interestingly, climate influenced the spatial patterns of vegetation and soil C density via different routes, having a positive effect on vegetation C density and a negative effect on soil C density. This estimation of the potential for increasing forest C storage provided new insights into the vital roles of China's forest ecosystems in future C sequestration. More importantly, our findings emphasize that climate constraints on forest C sequestration should be considered in reforestation strategies in China because the effects of climate were the opposite for spatial patterns of C density in vegetation and soil.

A high resolution lithospheric magnetic field model over China

CHAMP satellite data and ground-based magnetic observations are used and combined to map the lithospheric magnetic field over China by means of the revised spherical cap harmonic analysis(R-SHCA)modeling technique.The magnetic field is described to a spatial resolution of 150 km at the mean Earth’s radius,which represents a good compromise between the resolutions afforded by surface and satellite data.We compare the magnetic anomalies modeled at the regional scale with composite regions containing large-scale of tectonic structures.These regions,including the Tarim Basin and the Tibetan Plateau,are correlated with regional magnetic anomalies at satellite altitude but contain a significant number of small-scale and complex magnetic structures at the mean Earth’s radius.These magnetic anomalies are globally consistent with the known geological features in China but also offer a way to delineate the contours of the geological blocks and to discuss the connection between magnetic anomalies and the heat flow distribution in this region.